Incorrect file in zip archive for git repository

[roand@…]

Incorrect file in zip archive for git repository. In repository "test_git.tar.gz​" a file "A/a1.txt" of last commit must contains:

a1-1

A file "A/a1.txt" of zip archive for git repository contains:

a1

[Jun Omae]

Works for me. Please post how to reproduce it and what you expect if you have the issue.

[Jun Omae]

Okay. Reproduced it:

1. Extract test_git.tar.gz to /tmp
2. Execute repository add t12557 /tmp/test_git/.git git
3. Set /* to [browser] downloadable_paths
4. Visit /browser/t12557
5. Download a zip file via Zip Archive link in Download in other formats
6. A/a1.txt file contains a1 in the zip file, that should be a1-1

$ curl -s 'http://127.0.0.1/browser/t12557/?rev=add4fc1ec1d3237ad1cf585bb122c88fcefc23cc&format=zip' > /tmp/t12557.zip
$ unzip -l /tmp/t12557.zip
Archive:  /tmp/t12557.zip
  Length      Date    Time    Name
---------  ---------- -----   ----
        0  1980-01-01 00:00   A/
        2  2016-08-09 21:04   A/a1.txt
        4  2016-08-09 21:10   A/a2.txt
        0  1980-01-01 00:00   B/
        4  2016-08-09 21:05   B/b1.txt
        4  2016-08-09 21:12   B/b2.txt
---------                     -------
       14                     6 files
$ unzip -p /tmp/t12557.zip A/a1.txt | od -c
0000000   a   1
0000002

However, the A/a1.txt file in zip file downloaded from A directory is correct.

$ curl -s 'http://127.0.0.1/browser/t12557/A?rev=add4fc1ec1d3237ad1cf585bb122c88fcefc23cc&format=zip' > /tmp/t12557-A.zip
$ unzip -l /tmp/t12557-A.zip
Archive:  /tmp/t12557-A.zip
  Length      Date    Time    Name
---------  ---------- -----   ----
        4  2016-08-09 21:08   A/a1.txt
        4  2016-08-09 21:10   A/a2.txt
---------                     -------
        8                     2 files
$ unzip -p /tmp/t12557-A.zip A/a1.txt | od -c
0000000   a   1   -   1
0000004

[Jun Omae]

It seems this issue occurs when GitNode.get_entries() is recursively called via BrowserModule._iter_nodes(). The _iter_nodes() is used only in generating zip files.

>>> for node in BrowserModule(env)._iter_nodes(repos.get_node('')):
...   node.rev, node.created_rev, node.path
...
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', 'add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', u'')
('4b3ce925a1af49aca2b5f2397b9a35f8bc4c4185', '4b3ce925a1af49aca2b5f2397b9a35f8bc4c4185', u'A')
('7585aec50c0b0472e5a69b7f97bc045f9078c638', '7585aec50c0b0472e5a69b7f97bc045f9078c638', u'A/a1.txt')
('4b3ce925a1af49aca2b5f2397b9a35f8bc4c4185', '4b3ce925a1af49aca2b5f2397b9a35f8bc4c4185', u'A/a2.txt')
('f62fd985f264fa02f86605bd89f5fc3e7bd7941c', 'f62fd985f264fa02f86605bd89f5fc3e7bd7941c', u'B')
('cab4b75bbcdaf0b69e8f99feef07a7c63fe10b89', 'cab4b75bbcdaf0b69e8f99feef07a7c63fe10b89', u'B/b1.txt')
('f62fd985f264fa02f86605bd89f5fc3e7bd7941c', 'f62fd985f264fa02f86605bd89f5fc3e7bd7941c', u'B/b2.txt')

The created_rev attribute of node A/a1.txt should be eb818d752971e1bc5f8311bae49850a3f1d34da5, not 7585aec50c0b0472e5a69b7f97bc045f9078c638.

$ git log --oneline --no-abbrev-commit --name-status --graph --
*   add4fc1ec1d3237ad1cf585bb122c88fcefc23cc Merge branch 'A'
|\
| *   940380b367d996b384b82315ea581e13e0c55b18 Merge branch 'B' into A
| |\
| | * eb818d752971e1bc5f8311bae49850a3f1d34da5 Changed a1
| | | M A/a1.txt
| * | 4b3ce925a1af49aca2b5f2397b9a35f8bc4c4185 Changed a2
| | | M A/a2.txt
* | | f62fd985f264fa02f86605bd89f5fc3e7bd7941c Changed b2
| |/
|/|
| |   M B/b2.txt
* | bd7ecb85de78c7790f942355da3a619676a694e5 Changed b2
| | M   B/b2.txt
* | cab4b75bbcdaf0b69e8f99feef07a7c63fe10b89 Changed b1
|/
|   M   B/b1.txt
* 7585aec50c0b0472e5a69b7f97bc045f9078c638 First commit
  A     A/a1.txt
  A     A/a2.txt
  A     B/b1.txt
  A     B/b2.txt

Please try the following patch. I'll add unit test for this issue.

tracopt/versioncontrol/git/git_fs.py

@@ -645 +645 @@
             rev = repos.normalize_rev(to_unicode(rev))
         else:
             rev = repos.youngest_rev
+        created_rev = rev
 
         kind = Node.DIRECTORY
         p = path.strip('/')
@@ -662 +663 @@
             self.fs_perm, k, self.fs_sha, self.fs_size, fname = ls_tree_info
 
             # fix-up to the last commit-rev that touched this node
-            rev = repos.git.last_change(rev, p, historian)
+            created_rev = repos.git.last_change(rev, p, historian)
 
             if k == 'tree':
                 pass
@@ -677 +678 @@
                                   "kind '%(kind)s')", kind=k))
 
         self.created_path = path
-        self.created_rev = rev
+        self.created_rev = created_rev
 
         Node.__init__(self, repos, path, rev, kind)
 

After the patch:

>>> for node in BrowserModule(env)._iter_nodes(repos.get_node('')):
...   node.rev, node.created_rev, node.path
...
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', 'add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', u'')
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', '4b3ce925a1af49aca2b5f2397b9a35f8bc4c4185', u'A')
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', 'eb818d752971e1bc5f8311bae49850a3f1d34da5', u'A/a1.txt')
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', '4b3ce925a1af49aca2b5f2397b9a35f8bc4c4185', u'A/a2.txt')
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', 'f62fd985f264fa02f86605bd89f5fc3e7bd7941c', u'B')
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', 'cab4b75bbcdaf0b69e8f99feef07a7c63fe10b89', u'B/b1.txt')
('add4fc1ec1d3237ad1cf585bb122c88fcefc23cc', 'f62fd985f264fa02f86605bd89f5fc3e7bd7941c', u'B/b2.txt')

[Jun Omae]

Proposed changes in [7320f2e28/jomae.git].

[roand@…]

The patch works. Thank you for operational patch!

[Jun Omae]

Committed in [15044] and merged in [15045-15046].